ESMA, or the European Securities and Markets Authority, is the institution responsible for the ESEF format, i.e., the European standard for files in which the financial statements of listed companies are published. ESMA will also develop, based on a draft provided by EFRAG, a standard for tagging information in sustainability reports. Work on its implementation has been suspended. There is no point in issuing XBRL tagging schemes for \u201eold\u201d ESRSs if simplified standards will soon be in force.<\/p>\n\n\n\n
ESMA used the time saved for internal research, among other things. The aim was to check whether various AI-based tools could replace the labeling of information in reports. The conclusion is rather clear: they cannot. None of the models tested (both the latest general models and special models trained for this purpose were tested) were able to present accurate data selected from reports in which the data was not properly tagged. The number and, above all, the nature of the errors were too great for the data obtained in this way to be of any value to users. However, most tools performed quite well when presented with reports tagged with XBRL tags for analysis.<\/p>\n\n\n\n
The result of the study was fairly easy to predict. Large language models perform well at summarizing text content, but significantly worse with numbers, especially when they need to be understood in a specific context. After all, we don't want investors to receive information that the Gender Pay Gap index in our company has worsened by 30 percentage points, when we wrote in the report that it had improved by 3. Such seemingly minor errors (a change in the scale of values or attributes) can influence investors' decisions regarding our company.<\/p>\n\n\n\n
On the other hand, data tagging does not have to be laborious. I would compare it to whether we use styles when writing a Word document or manually change all text parameters (font style, adjustment, size, colors) in each paragraph. Some of the reporting support tools already available on the Polish market today enable automatic tagging of reports as they are being created.<\/p>\n\n\n\n
I expect that tagging reports will remain mandatory. This will enable the creation of reports that can be easily analyzed with precision, whether by humans or with the help of AI tools.<\/p>","protected":false},"excerpt":{"rendered":"
Tagging data with XBRL tags was supposed to be a nightmare that would affect all companies preparing sustainability reports. The confusion caused by Omnibus and the revision of ESRS standards has postponed the problem, but eventually it will have to be addressed. However, is it worth doing so in an era of rapid development of AI tools? Perhaps artificial intelligence will be able to handle it and we will not have to deal with data tagging?<\/p>","protected":false},"author":2,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"footnotes":""},"categories":[13],"tags":[19,20],"class_list":["post-1388","post","type-post","status-publish","format-standard","hentry","category-newsletter-piotra-biernackiego","tag-csrd-esrs","tag-raportowanie-esg"],"yoast_head":"\n